Cindicator is developing a financial management platform based on “hybrid intelligence” – a combination of analysts’ insights and machine learning models – so as to enable investors to manage their capital more efficiently and analysts to monetize their intellectual assets, while providing the tools and data necessary for making investment decisions under conditions of market volatility and uncertainty.
Cindicator uses artificial intelligence to synthesize the collective intelligence of forecasters into conclusive signals. Analysts on the platform generate various forecasts daily by answering a number of specific questions about the price levels of different financial assets, macroeconomic indices, and events significantly influencing the market. Although it isn’t clear how the questions posed to the community are selected, examples given in the whitepaper include:
- Create a forecast of the minimum and maximum price levels of Bitcoin for the coming seven-day period.
- Will the Tesla stock price surge to $345 during market hours on Friday?
- Will the U.S. unemployment rate be greater than or equal to 4.5%, according to the 2 June report?
- Will Bancor collect more than 100 million during the first week of ICO?
- What is the probability of Trump’s impeachment during the next three months?
The platform then uses a machine learning ensemble (cleaning, clustering methods, linear regressions, Bayesian models, xgboost on decision trees, genetic algorithms, and neural networks) to dynamically calculate weights for individual forecasters, identify and correct systematic errors, eliminate noise, and generate final predictions and trading signals for use by traders. Forecasters receive monthly rewards in accordance with the amount of their activity and the accuracy of their predictions.
Cindicator’s current mobile app provides an interface for forecasters that makes it possible to generate a new data point within a few minutes. As of now, the Android app has 5000-10000 installs and is rated an average of 4.6 among 213 reviewers. The platform adapts dynamically, continuously correcting its forecaster weights, and employs trading robots which complete a real or model trade linked to every question posed. Confidence weights take into account the forecaster’s personal track record in terms of accuracy and types of forecasts, dynamic feedback following each trade, and a predictive model capable of identifying superforecasters in the group. The trading robots are used to test various trading strategies and hypotheses, and to run retrospective or predictive tests.
Cindicator’s CND tokens will provide their holders with differing levels of access to platform services, depending on the quantity of CNDs in their possession. Token holders will be able to access market indicators, indices, and generated sentiments, as well as auxiliary trading products (Telegram and Slack bots, notiers, portfolio monitoring) and analytical products (ICO ratings, market condition analysis, ICO due diligence, investor portfolio analysis). The access levels are still undetermined, to be formulated based on the results of the crowdsale, market dynamics, and Cindicator’s internal economy.
According to the whitepaper, the platform currently consists of:
- A business logic module including a backend with basic business logic, an administrative system, mobile applications, and an under-development web application.
- A prediction module including data acquisition and filtration, feature extraction, formation of hypotheses and mathematical models, validation and optimisation of parameters for predictive models, and synthesis of accurate predictions.
- A trading module including input from the predictive module, integration with exchanges and processing of resulting data, backtests and forward tests for parameters of trading strategies, and implementation of trading strategies through trading robots.
Once development is complete, the platform will deliver daily/weekly/monthly distribution of indicators via messenger/email, a SaaS dashboard with access to indicators and analytics, a mobile application, and API access. Nevertheless, certain parts of the infrastructure which are intended for use directly in capital management (by traders’ teams, machine learning models, and trading strategies) will not be openly accessible but remain centralized.
The machine learning employed by the platform is based on studying and identifying the behavioral patterns of forecasters, conducting experiments with groups and clusters, experimenting with predictive models, and time series analyses of the market and forecaster predictions. The predictive models are based either on particular (superforecaster) clusters or on all forecasters, and the mathematical models are based on the theories of phase transitions and game theory, using also fractal geometry to forecast critical points. Parameters for each model are tuned with regards to each financial asset, and every model constantly learns on the basis of new data. Accuracy and quality are assessed via back-testing, using both standard scores (RMSE, ROC, MAE, Pearson) and intrinsic evaluation functions.
In order to validate and develop its platform, Cindicator plans to use part of the crowdsale funds to create a trading portfolio with active cryptotrading, protective buy and hold of crypto assets, and active trading of traditional financial assets. The necessary legal structure and licenses for this are to be prepared after the token sale, also using a part of the collected funds. The portfolio will be managed by the company’s traders and bots, utilizing the data, signals, and analytics obtained through the company’s technology. Various strategies will be applied, from short-term trades to long-term investments.
To form a steadily growing internal economy, Cindicator will create an internal motivational pool of CND tokens, used to encourage forecasters and other contributors. Tokens will be funneled to the pool primarily from the selling of Cindicator analytical products. It is not clear which products will be available to token holders, and which will be sold for tokens. The company will form an additional motivational ETH/BTC pool based on the trading portfolio, so as to tie compensation to performance. Profits from the portfolio will be partly reinvested, partly distributed to the Cindicator team, and partly funneled to this motivational pool, although in what percentages is not yet known. The whitepaper also describes a process wherein forecasters will be given publicly available personal ratings based on accuracy, and each month, the top 2% will share a predetermined cash prize (presumably from the motivational pool, or another fund).
Cindicator’s final objective is to create an infrastructure for investment funds, which will purchase access to the platform via CND tokens and pay a performance fee from their profits via funneling to the motivational pool. The number of investment funds with access to the full infrastructure will be limited in order to maximize efficiency. The company previously launched test integrations with hedge funds and banks, but realized that selling to a large number of B2B customers would be unwise from a business perspective. The infrastructure is scheduled to be available for funds in 2019, and access will be granted only to those who own a significant (yet undetermined) number of CND tokens.
Cindicator was founded in 2015 by CEO Mike Brusov, CTO Yuri Lobyntsev, and COO Artem Baranov, with the overall goal of developing hybrid intelligence comprised of crowd wisdom and AI/ML. Mike Brusov was previously cofounder of Wobot, a Moscow-based developer of a professional social media monitoring and analytics tool, as well as head of mobile programmatic products at Between Digital, a Russian RTB-platform for publishers. Yuri Lobyntsev was previously CEO and cofounder of Octabrain, a software company in the field of neurointerfaces and neural networks, as well as CEO and founder of Oumobile, a mobile application studio, both of which were headquartered in Saint-Petersburg. Artem Baranov was previously COO at Octabrain, and before that at Crimson Jackets, a Moscow-based creative studio for digital products. Next to join the team was data scientist and backend developer Alexander Frolov, who previously worked in structural bioinformatics at Biocad pharmaceuticals. Following the decision to focus on finances, the team expanded to include head of analytics Kate Kurbanova and CIO Nodari Kolmakhidze, both experienced traders. According to the website, the team now includes seven additional developers as well as 15 advisors and partners including Charlie Shrem (COO at Jaxx and founder at Bitcoin Foundation), Anthony Diiorio (CEO and founder and Jaxx and Decentral, founder at Ethereum), Evan Cheng (Director of Engineering at Facebook), and Reese Jones (associate founder at Singularity University).
The company released the first iOS-based version of its platform in December 2015. From June to November 2016, the team worked on creating the first set of machine learning models, improving the collective intelligence system, and developing various trading strategies. In January 2017, an API for trading signals was launched, enabling test integrations with numerous hedge funds and banks in the following four months, and from November 2016 to March 2017, Cindicator took part and top-ranked in the first batch of the Moscow Exchange fintech incubator. The company has been granted $120,000 from Microsoft and became a member of the BizSpark program, and in April-May 2017 raised $200,000 from a number of fintech investors in a seed venture round. (According to their Crunchbase entry, the company has raised $500,000 in seed investment).
To date, over 8,000 forecasters have made 230,000 forecasts on the platform. The company initially focused on binary/probabilistic and price-range related questions, but has been testing other promising signals including questions regarding single price levels, buying and selling of crypto-assets, and ranking of ICOs according to the degree of their probable success or trustworthiness. The artificial intelligence system has been aggregating the results from all forecasters and determining conclusive signals, which have then been used to trade a model portfolio. The results, which are comprehensively presented in the whitepaper, look promising. For example:
- From July 2016 to February 2017, the model portfolio increased by 123% using binary/probabilistic signals, and in 148 trading days between October 2016 and March 2017, managed to reach a return of 11.68% (annual profitability of 19.8%) by using a simple counter-trend strategy (purchase at min and sell at max) based on signals from price-range related questions – both on traditional markets.
- In January 2017, Cindicator launched a public experiment with the Moscow Exchange (MOEX), involving 925 unrelated participants, 40% of whom had never made trades on the exchange before. The company received 25,000 forecasts regarding min/max levels of four financial assets (USD/RUB futures, Brent oil futures, silver futures, and gold futures) for three weeks on a daily basis, excluding weekends and holidays. The robot aggregated the forecasts into a signal (including entry level, stop loss, and take profit) and simulated the trades. Entry levels were posted to the company’s public Telegram channel on a daily basis prior to market opening. Based on these forecasts, Cindicator’s robot modelled 57 trades over the course of the experiment, 36 of which were profitable. The completed trades showed a 3% return in 29 days (equivalent to 26% annual).
Cindicator’s token sale is being conducted via WhiteList registration in four tiers, where each tier differs in min/max individual contribution and cap, but not in token price – there are no bonuses or discounts. This is intended to provide a measure of security and fairness, and to foster the growth of a dedicated community of forecasters and contributors.
*Disclosure – At the time of writing, CryptoRated‘s staff owned no Cindicator tokens.